Modern and Emerging Frontiers in Deep Learning
for Security and Defense

10:30 am
Friday December 8th, 2023
Room 3107
Patrick F. Taylor Hall




In today's security and defense environment, covering land, sea, air/space, and cyberspace, we face unprecedented complexity, interconnection, and scale. This reality presents unique challenges in detecting, characterizing, and neutralizing potential threats, emphasizing the need to go beyond human capabilities alone. In this presentation, I will discuss these challenges and share my innovative approaches using modern and emerging deep learning technologies. Specifically, I'll delve into a past project where I designed a custom deep neural autoencoder architecture, employing unsupervised learning to identify subtle signals in intricate, noisy sensor data. Moving forward, we'll explore neuromorphic computing—an emerging field related to deep learning, aiming to create energy-efficient spiking neural networks inspired by biological systems. We'll discuss its applications in edge computing for security and defense, as well as its potential to address the growing power consumption concerns of current deep learning technology. Finally, I'll detail my current initiatives and future objectives in creating a comprehensive "AI for Security" research program at LSU. I'll delve into specific plans, including a discussion on utilizing Large Language Models (LLMs) in cyberspace.

James M. Ghawaly

James Ghwaly

SDMI, Louisiana State University

Dr. James Ghawaly is a Senior Research Scientist - AI/ML in the Louisiana State University Stephenson Disaster Management Institute (SDMI), where he is leading efforts to develop applied artificial intelligence and machine learning techniques. Prior to LSU, he was a research data scientist at the Department of Energy Oak Ridge National Laboratory, where he led and supported a variety of federally sponsored research efforts to develop machine learning approaches for national security challenges. Dr. Ghawaly specializes in the utilization of deep learning methodologies across diverse domains, harnessing the power of AI and machine learning to enhance global security, safety, and overall well-being. He has pioneered innovative and state-of-the-art models for autonomously analyzing complex and noisy sensor data. He leverages foundational large language models (LLM) to develop applications that extract valuable information and insights from unstructured textual data, including human-generated content. Additionally, he has spearheaded research initiatives in neuromorphic computing, crafting neural network models that demonstrate notably superior energy efficiency and computational power when juxtaposed with traditional deep learning approaches.